LoadLeveler Advance Reservation plugin
Commands
- llmkres - make a reservation
- llqres - query a reservation
- llrmres - cancel a reservation
ARDL tags mapped on llmkres flags
TimeWindow | StartTime: -t [mm/dd[/[cc]yy]] HH:MM EndTime/Duration: -d duration of the reservation in minutes |
ReservedSlotsCount AdditionalReservedSlotsCount IndividualPhysicalMemory IndividualSwapMemory | -n number of nodes to reserve -f path to a job command file that will be submitted and the first job step used to determine what resources to reserve (mutually exclusive with -h) |
CandidateHosts | -h blank-delimited list of machines to reserve (mutually exclusive with -n) |
UserName | -U blank-delimited list of users who can use the reservation |
GroupName | -G blank-delimited list of LoadLeveler groups whose users can use the reservation |
CPUArchitecture ExclusiveExecution FileSystem NativeResource OperatingSystem Project ReservationName | - |
Grid Monitoring
HostInfo
HostInfo members | LoadLeveler properties - llstatus (-f|-r) (category) |
---|---|
HostName | Name (%n) Machine name |
CpuArch | Default to powerpc |
OSName | Default to linux |
State | Startd (%sts) State of the startd daemon described below |
CpuCount | Cpus (%cpu) Number of CPUs |
UsedCpuCount | |
ReservedCpuCount | |
CpuSpeed | |
CpuUtilization | LdAvg (%l) Berkeley one-minute load average - is it? |
QueueLength15S | |
QueueLength1M | |
QueueLength15M | |
PagingActivity | |
FreeMemory | |
FreeSwap | FreeVMemory (%v) Available swap space |
FreeFilesystem | FreeDisk (%d) Available disk space in the LoadLeveler execute directory |
TotalMemory | Memory (%m) Physical memory |
TotalSwap | |
TotalFilesystem | |
Jobs | querying getJobData |
Queues | |
AdditionalResources |
Node States mapping
State of the startd daemon | HostInfo.State |
---|---|
Busy Idle Run (Running) | up |
Down | down |
Drned (Drained) Drning (Draining) Flush None Suspnd (Suspend) | offline |
JobInfo
JobInfo members | LoadLeveler properties - llq (-f|-r) (category) |
---|---|
JobID | Step ID (%id) |
JobName | Job name (%jn) |
OwnerName | Job owner (%o) |
JobStatus | Status (%st) described below |
QueueName | Class (%c) |
ReservationID | Reservation ID (%R) |
Description | |
SubmissionTime | Queue date (%dq) |
StartTime | Dispatch Date (%dd) |
EndTime | Completion date (%dc) |
EstStartTime | |
EstEndTime | |
ExecutionHosts | Host name (%h) |
MemUsage | |
CpuTime |
Job States mapping
Job state in LoadLeveler | JobInfo.JobStatus |
---|---|
Deferred (D) Idle (I) Not Queued (NQ) Not Run (NR) Pending (P) Rejected (X) Reject Pending (XP) Remove Pending (RP) Starting (ST) System Hold (S) User & System Hold (HS) User Hold (H) | Pending |
Checkpointing (CK) Complete Pending (CP) Preempt Pending (EP) Resume Pending (MP) Running (R) | Running |
Preempted (E) Vacated (V) Vacate Pending (VP) | Suspended |
Canceled (CA) Completed (C) Removed (RM) Terminated (TX) | Done |
QueueInfo
QueueInfo members | LoadLeveler properties - llclass -l |
---|---|
QueueName | Name |
QueuePriority | Priority |
QueueEnabled | always true |
MaxJobs | Maximum_slots |
PendingJobs RunningJobs JobIDs | querying getJobData |
IsDefault | Name == 'No_Class' (configuration dependent) |
Walltime | Wall_clock_limit |
AdditionalInfo |
ReservationInfo
ReservationInfo members | LoadLeveler properties - llqres (-l|-r) |
---|---|
ReservationName | ID |
Type | always 'ar' |
StartTime | Start Time |
EndTime | Expected End Time |
Owner | Owner |
Resources |
JSDL
JSDL markups | LoadLeveler job command file keywords |
---|---|
JobIdentification | |
JobName | @job_name |
JobAnnotation | |
JobProject | @account_no |
Description | @comment |
Resources | |
CandidateHosts | @requirements = ((Machine == { "HostName1" "HostName2" }) |
FileSystem | |
ExclusiveExecution | @node_usage = not_shared |
OperatingSystem | |
CPUArchitecture | |
IndividualCPUSpeed | |
IndividualCPUTime | @cpu_limit |
IndividualCPUCount | @tasks_per_node |
IndividualNetworkBandwidth | |
IndividualPhysicalMemory | @requirements = (Memory > n) |
IndividualVirtualMemory | |
IndividualDiskSpace | @requirements = (Disk > n) |
TotalCPUTime | @job_cpu_limit |
TotalCPUCount | @total_tasks |
TotalPhysicalMemory | |
TotalVirtualMemory | |
TotalDiskSpace | |
TotalResourceCount | @node |
!Reservation | |
NativeSpecification | |
Queue | @class |
NodeProperties | |
ConsumableResources | |
JobCategory |